As electronics evolve, there is an increased need for miniature switches that are provided on semiconductor substrates along with other semiconductor components to form various types of circuits. These miniature switches often act as relays, and are generally referred to as micro-electro-mechanical system (MEMS) switches. MEMS switches generally include a moveable portion such as a cantilever, which has a first end anchored to the semiconductor substrate, and a second free end having a cantilever contact. When the MEMS switch is activated, the cantilever moves the cantilever contact against a substrate contact on the semiconductor substrate and under the cantilever contact.
Turning to FIGS. 1A and 1B, a semiconductor device 10 having a MEMS switch 12 is illustrated. The MEMS switch 12 is effectively formed on a semiconductor substrate 14. The MEMS switch 12 includes a cantilever 16, which is formed from a conductive material, such as gold. The cantilever 16 has a first end and a second end. The first end is coupled to the semiconductor substrate 14 by an anchor 18. The first end of the cantilever 16 is also electrically coupled to a first conductive pad 20 at or near the point where the cantilever 16 is anchored to the semiconductor substrate 14. Notably, the first conductive pad 20 may play a role in anchoring the first end of the cantilever 16 to the semiconductor substrate 14 as depicted.
The second end of the cantilever 16 forms or is provided with a cantilever contact 22, which is suspended over a contact portion 24 of a second conductive pad 26. Thus, when the MEMS switch 12 is actuated, the cantilever 16 moves the cantilever contact 22 into electrical contact with the contact portion 24 of the second conductive pad 26 to electrically connect the first conductive pad 20 to the second conductive pad 26.
To actuate the MEMS switch 12, and in particular to cause the second end of the cantilever 16 to move the cantilever contact 22 into contact with the contact portion 24 of the second conductive pad 26, an actuator plate 28 is disposed over a portion of the semiconductor substrate 14 and under the middle portion of the cantilever 16. To actuate the MEMS switch 12, a potential difference is applied between the cantilever 16 and the actuator plate 28. The presence of this potential difference creates an electrostatic force that effectively moves the second end of the cantilever 16 toward the actuator plate 28, thus changing the position of the cantilever 16 from the position illustrated in FIG. 1A to the position illustrated in FIG. 1B.
Typically, the first conductive pad 20, the second conductive pad 26, and the actuator plate 28 are formed from a single metallic or conductive layer, such as gold, copper, platinum, or the like. The particular form factor for the first conductive pad 20, second conductive pad 26, and actuator plate 28 is provided through an etching or other patterning technique. With continued reference to FIGS. 1A and 1B, the MEMS switch 12 may be encapsulated by one or more encapsulating layers 30 and 32, which make up a wafer level package (WLP) around the MEMS switch 12. Moreover, the encapsulating layers 30 and 32 form a substantially hermetically sealed cavity about the cantilever 16. The cavity is generally filled with an inert gas. Once the encapsulation layers 30 and 32 are in place and any other semiconductor components are formed on the semiconductor substrate 14, a plastic overmold 34 may be provided over the encapsulation layers 30 and 32 and any other semiconductor components.
With continued reference to FIGS. 1A and 1B, the substrate 14 is preferably formed using a semiconductor-on-insulator process, such as a silicon- or sapphire-on-insulator process. In particular, the substrate 14 includes a handle wafer 36 that is formed from silicon, sapphire, glass, or like material to form a foundation layer for the semiconductor device 10. The handle wafer 36 is typically a few hundred microns thick. An insulator layer 38 is formed over the handle wafer 36. The insulator layer 38 is generally formed from an oxide, such as Silicon Dioxide (SiO2), which may range in thickness from 0.1 to 2 microns in the preferred embodiment. A device layer 40, which may include one or more layers, is formed using an appropriate semiconductor material.
The device layer 40 is the layer or layers in which active semiconductor devices, such as transistors and diodes that employ PN junctions, are formed. The device layer 40 is initially formed as a base semiconductor layer that is subsequently doped with N-type and P-type materials to form the active semiconductor devices. Thus, the active semiconductor devices, except for any necessary contacts or connections traces, are generally contained within the device layer 40. Those skilled in the art will recognize various techniques for forming active semiconductor devices in the device layer 40. A metal-dielectric stack 42 is formed over the device layer 40, wherein a plurality of metal and dielectric layers are alternated to facilitate connection with and between the active devices formed in the device layer 40. Further, in the preferred embodiment the handle wafer 36 is made of a high-resistivity semiconductor material where resistance is greater than 50 ohm-cm.
With the present disclosure, active semiconductor devices may be formed in the device layer 40 and connected to one another via the metal-dielectric stack 42 directly underneath the MEMS switch 12. Since the device layer 40 resides over the insulator layer 38, high voltage devices, which may exceed ten (10) volts in operation, may be formed directly under the MEMS switch 12 and connected in a way to control operation of the MEMS switch 12 or associated circuitry. Although silicon is described in the preferred embodiment, the semiconductor material for the device layer 40 may include gallium arsenide (GaAs), gallium nitride (GaN), indium phosphide (InP), silicon germanium (SiGe), sapphire, and like semiconductor materials. The device layer 40 typically ranges in thickness from 0.1 microns to 20 or more microns.
As illustrated in FIGS. 1A and 1B, a passivation layer 44 may be provided over the metal-dielectric stack 42. As may be best seen from the perspective view of FIG. 2, a metal layer used to form the first conductive pad 20, the second conductive pad 26, and the actuator plate 28 may be formed over the passivation layer 44 and etched to form the respective ones of the first conductive pad 20, the second conductive pad 26, and the actuator plate 28. Prior to packaging, the cantilever 16 is ‘released’ and is free to actuate or deform. In particular, the cantilever 16 may be released following formation of a small micro-cavity surrounding the MEMS switch 12. A sacrificial material such as polymethylglutarimide (PMGI) is etched away using wet etches. Following drying and cleaning of the MEMS switch 12, a dielectric is used to hermetically seal the micro-cavity. The deposition temperature for the dielectric is typically 250° C. Later in the manufacturing process, the device can experience multiple exposures to 260° C. solder reflow during attachment of a module incorporating the MEMS switch 12 to an end-user laminate.
A problem of undesirable deformation of the MEMS switch 12 often occurs due to a significant difference in the coefficient of thermal expansion (CTE) between the metal comprising MEMS switch 12 and the semiconductor or insulator comprising passivation layer 44. The CTE of the metal making up the MEMS switch 12 often ranges from two to seven times larger than the CTE of the semiconductor or insulator making up the passivation layer 44. At room temperature (i.e., 25° C.), the difference in the CTE does not present a problem. However, during manufacture, assembly, or operation of the MEMS switch 12, the temperature of the MEMS switch 12 and the substrate 14 (FIGS. 1A and 1B) including the passivation layer 44 can range from 85° C. to 400° C. In such circumstances, particularly in the case of an ohmic contact switch function for MEMS switch 12, it is desirable to ensure that the cantilever contact 22 and second conductive pad 26 or passivation layer 44 do not make contact. FIG. 3 illustrates how differences in CTE may lead to a thermally induced deformation of the cantilever 16. A plurality of dots shown in a side cross-section of the cantilever 16 and the anchor 18 represent individual metal domains making up the cantilever 16 and the anchor 18. As the MEMS switch 12 is heated during manufacturing and/or assembly, the metal domains expand and push against each other. The domains further from the passivation layer 44 are allowed to expand more than domains closer to the passivation layer 44, thereby producing a deflection force on the cantilever 16. The deflection force is represented by an arrow at the free end of the cantilever 16. In this case, the deflection force urges the cantilever towards the passivation layer 44.
FIG. 4 depicts the results of a finite element simulation of the mechanical effects experienced by the MEMS switch 12 when the MEMS switch 12 is heated to a steady state temperature of 300° C. The finite element simulation shows when the MEMS switch 12 reaches a temperature of 300° C., the cantilever 16 will have rotated enough that the cantilever contact 22 will be in contact with the second conductive pad 26. MEMS switch 12 has a switch open state that typically maintains a one-half micrometer gap between the cantilever contact 22 and the second conductive pad 26. Further finite element simulations show that the deflection of the cantilever 16 may allow the cantilever contact 22 to traverse gap distances that exceed one-half micrometer.
Notice that a rotational axis 46 of the cantilever 16 is perpendicular to a longitudinal axis 48 of the cantilever 16. As suggested by the finite element simulations, due to the combination of the significant difference in CTE between the metal of MEMS switch 12 and the semiconductor or insulator comprising passivation layer 44 and the elevated temperatures experienced by the MEMS switch 12 during manufacturing, assembly, or operation, the cantilever 16 may be thermally deflected to rotate about the rotational axis 46. As the temperature of the MEMS switch 12 increases, the rotation of cantilever 16 may become so pronounced that the cantilever contact 22 will contact the second conductive pad 26. An adhesion between the cantilever contact 22 and the second conductive pad 26 may prevent the cantilever contact 22 and the second conductive pad 26 from breaking contact as the temperature of the MEMS switch 12 decreases. A failure to break contact between the cantilever contact 22 and the second conductive pad 26 will result in a failed MEMS switch along with a failed product incorporating the MEMS switch 12.
FIG. 5 depicts a MEMS switch 50 that has been proposed by the Defense Advanced Research Projects (DARPA). The DARPA MEMS switch 50 includes a substrate 52 onto which a first conductive pad 54 and a second conductive pad 56 are formed. A conductive anchor 58 is coupled to the first conductive pad 54A conductive cantilever disc 60 has a first end portion 62, a middle portion 64, and a second end portion 66. The second end portion 66 of the conductive cantilever 60 is suspended over the second conductive pad 56 with the middle portion 64 being between the first end portion 62 and the second end portion 66. A conductive actuator plate 68 is formed onto the substrate 52 at a location beneath the middle portion 64 of the conductive cantilever disc 60 and between the first conductive pad 54 and the second conductive pad 56. When a potential difference is applied to the actuator plate 68, an electrostatic force pulls the cantilever disc 60 towards the substrate 52 until an electrical contact 70 comes into contact with the second conductive pad 56.
Unfortunately, it has been found that the DARPA proposed MEMS switch 50 suffers from yield losses comparable to those experienced by MEMS switch 12 (FIGS. 1A, 1B, 2, 3 and 4) when subjected to the elevated temperatures experienced by the MEMS switch 50 during manufacturing, and assembly. To demonstrate the thermal actuation occurs, the DARPA proposed MEMS switch 50 shown in FIG. 6 was modeled with a thermal finite element analysis. A cantilever disc 72 attached to conductive anchors 74 and 76 was simulated under a manufacturing thermal environment of 300° C. An attachment axis 78 for anchor 74 and an attachment axis 80 for anchor 76 are tangential to a circumference around the cantilever disc 72. Therefore, a deflection of the cantilever disc 72 is expected as the cantilever disc 72 expands due to a high thermal environment of 300° C. during manufacturing. In fact, a simulation result predicts that a tip 82 of the cantilever disc 72 would deflect towards a substrate 84 by at least 0.3 micrometers (μm). Such a large tip deflection could potentially result in manufacturing yield loss or shifts in critical device parameters, as well as product failure of any product relying on MEMS switch such as the DARPA proposed MEMS switch 50 (FIG. 5).
Significant yield loss, which may approach upwards 50%, may be attributed to this thermally induced actuation during manufacture of devices with the kinds of attachment configurations described above. Thus, the need for a structure which can prevent this kind of thermal actuation is apparent.